The Stroop task, considered by many to be a paradigmatic measure of selective attention, has often been employed to investigate attention deficits in schizophrenia. Card and single-trial versions of this task have yielded different results. In this study both card and single-trial versions were administered to healthy controls (n = 24) and patients with schizophrenia (n = 55). No differences in reaction time (RT) interference were found on either version. On the single-trial version, patients showed greater RT facilitation and error rate interference, evidence for a deficit in selective attention. Methodologic and analytic issues that account for the mixed results from earlier card Stroop studies are addressed. It is concluded that single-trial versions provide greater sensitivity to selective attention pathology in schizophrenia.